Handwritten-numeral-recognition

This is a project about using CNN to realize handwritten numeral recognition

How to use

Write each number from 0 to 9 a hundred times to prepare the data. Then run get_picture.py to capture these numbers into computer.

Then run convert_data_*.py to compress the data and make the data. The train_data.npy or train_data.tfrecords should be generaterd.

Run train_v*.py to train te model.

Run eval_v*.py to get the result.

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
README.md		README.md
convert_data_TFRecord.py		convert_data_TFRecord.py
convert_data_numpy.py		convert_data_numpy.py
eval_v1.py		eval_v1.py
eval_v2.py		eval_v2.py
eval_v3.py		eval_v3.py
eval_v4.py		eval_v4.py
eval_v5.py		eval_v5.py
get_picture.py		get_picture.py
train_v1_compressed_TFRecord.py		train_v1_compressed_TFRecord.py
train_v2_compressed_Numpy.py		train_v2_compressed_Numpy.py
train_v3_compressed_Numpy.py		train_v3_compressed_Numpy.py
train_v4_origin_Numpy.py		train_v4_origin_Numpy.py
train_v5_origin_Numpy.py		train_v5_origin_Numpy.py